Taming Numbers and Durations in the Model Checking Integrated Planning System

Abstract

The Model Checking Integrated Planning System (MIPS) is a temporal least commitment heuristic search planner based on a flexible object-oriented workbench architecture. Its design clearly separates explicit and symbolic directed exploration algorithms from the set of on-line and off-line computed estimates and associated data structures. MIPS has shown distinguished performance in the last two international planning competitions. In the last event the description language was extended from pure propositional planning to include numerical state variables, action durations, and plan quality objective functions. Plans were no longer sequences of actions but time-stamped schedules. As a participant of the fully automated track of the competition, MIPS has proven to be a general system; in each track and every benchmark domain it efficiently computed plans of remarkable quality. This article introduces and analyzes the most important algorithmic novelties that were necessary to tackle the new layers of expressiveness in the benchmark problems and to achieve a high level of performance. The extensions include critical path analysis of sequentially generated plans to generate corresponding optimal parallel plans. The linear time algorithm to compute the parallel plan bypasses known NP hardness results for partial ordering by scheduling plans with respect to the set of actions and the imposed precedence relations. The efficiency of this algorithm also allows us to improve the exploration guidance: for each encountered planning state the corresponding approximate sequential plan is scheduled. One major strength of MIPS is its static analysis phase that grounds and simplifies parameterized predicates, functions and operators, that infers knowledge to minimize the state description length, and that detects domain object symmetries. The latter aspect is analyzed in detail. MIPS has been developed to serve as a complete and optimal state space planner, with admissible estimates, exploration engines and branching cuts. In the competition version, however, certain performance compromises had to be made, including floating point arithmetic, weighted heuristic search exploration according to an inadmissible estimate and parameterized optimization.

Organizational commitment and its effects on organizational citizenship behavior in a high-unemployment environment

“…Although in temporal planning the most important criterion for quality is the makespan, we will also consider the number of actions of the plan as an additional criterion for the plan quality. We have compared the quality of the plans of TPSYS-LC with other (both domainindependent and domain-dependent) temporal planners, such as LPG1.0 11 (Gerevini & Serina, 2002a), Mips 12 (Edelkamp, 2002), SHOP2 (Nau, Cao, Lotem, & Muñoz Avila, 1999;Nau, Muñoz Avila, Cao, Lotem, & Mitchell, 2001), TALPlanner (Kvarnstrom & Doherty, 2000), TLPlan (Bacchus & Kabanza, 2000) and VHPOP (Younes & Simmons, 2002). We have rerun the problems again in both LPG1.0 and Mips, while, for the rest of planners, we have used the results provided in IPC-2002 13 .…”

“…If we focus on temporal planning, the last decade has seen many new works: O-PLAN (Currie & Tate, 1991), which integrates both planning and scheduling processes into a single framework, ZENO (Penberthy & Weld, 1994), IxTeT (Ghallab & Laruelle, 1994) and parcPLAN (El-Kholy & Richards, 1996), which cope with temporality on actions and temporal constraints, etc. Graphplan success has allowed the development of more modern temporal planners: TGP (Smith & Weld, 1999), whose ideas are very valuable in temporal environments, TP4 (Haslum & Geffner, 2001) and Sapa (Do & Kambhampati, 2001), which handle concurrent durative actions and use heuristic metrics to deal with resources in planning, LPG (Gerevini & Serina, 2002a) and Mips (Edelkamp, 2002). Both LPG and Mips showed an outstanding performance in the last International Planning Competition, demonstrating that heuristic and local search are very useful in planning and especially in temporal planning.…”

Deterministic Oversubscription Planning as Heuristic Search: Abstractions and Reformulations

“…In principle, the reduction of Keyder and Geffner (2009) from net-benefit to classical planning can be used to reduce OSP to classical planning with realvalued state variables (Koehler, 1998;Helmert, 2002;Fox & Long, 2003;Hoffmann, 2003;Gerevini, Saetti, & Serina, 2003;Gerevini et al, 2008;Edelkamp, 2003;Dvorak & Barták, 2010;Coles, Coles, Fox, & Long, 2013). So far, however, progress in heuristic-search classical planning with numeric state variables has mostly been achieved around direct extensions of delete relaxation heuristics via "numeric relaxed planning graphs" (Hoffmann, 2003;Edelkamp, 2003;Gerevini et al, 2003Gerevini et al, , 2008. Unfortunately, these heuristics do not preserve information on consumable resources such as budgeted operator cost in oversubscription planning: the "negative" action effects that decrease the values of numeric variables are ignored, possibly up to some special handling of so-called "cyclic resource transfer" (Coles et al, 2013).…”

Using the Context-Enhanced Additive Heuristic for Temporal and Numeric Planning

“…For this purpose, we use a transformation that is -at least similarly -also used by several other planning systems, e.g., by MIPS and LPG (Edelkamp 2003;Gerevini, Saetti, and Serina 2008): When compressing a condition triple, e.g. c ⊢ , c ↔ , c ⊣ , we remove all those persistent and end conditions that are made true by the associated operator op itself, i.e., all persistent and end conditions v = w for which op contains a start effect c, v = w (ignoring the effect condition c).…”